Skip to content
conftrace_

floating-point format

3 papers

Explore in graph

Co-occurring keywords · papers mentioning both

low-bit inference (1) neural gradient (1) neural processing unit (1) gradient sparsity (1) large language model (1) log-normal distribution (1) activation outlier (1) gradient quantization (1) inference efficiency (1) post-training quantization (1)

Papers

Unleashing Low-Bit Inference on Ascend NPUs: A Comprehensive Evaluation of HiFloat Formats ACL 2026 AMXFP4: Taming Activation Outliers with Asymmetric Microscaling Floating-Point for 4-bit LLM Inference ACL 2025 Neural gradients are near-lognormal: improved quantized and sparse training ICLR 2021